Glitch in US computer revolution: 'How do you talk to these things?'
Boston — For the contractor building Minneapolis's new Luthera Brotherhood Building, the good news was that the window glass arrived on schedule. The bad news was that each of the 12,000 panes was three inches too long, due to an error in a computer's "software."
In recent months, similar errors have wiped out a laboratory research project at the University of Southern California, and caused Austrialia's Federal Health Office to issue $250 million in overpayments.
Telling a computer to do something is like teaching a child long division. You have to start with the basics, and make each step perfectly clear. Even the most powerful data- processing device requires "software," a detailed step-by-step program of instructions, before it can perform useful tasks.
During the 1970s, the cost of computer hardware dropped to record lows, creating a demand for thousands of new data-processing applications. But software design remained a costly and tedious process, causing mistakes and delays that are threatening to slow the pace of America's computer revolution.
"People are realizing that a computer without software is nothing but a hunk of iron," says David Sturvetant of the Association of Data Processing Service Organizations.
One industry source estimates the average Fortune 1000 company has two years' worth of in-house software work waiting to be done.
Programmers, the scribes of the industry, are in great demand. The Department of Labor estimates that the US has 35,000 unfilled programming jobs.
And a study by Arthur D. Little Inc., a Cambridge, Mass., research firm, predicts that a software shortage will cost the American data processing industry $30 billion in lost revenues through 1984.
This bottleneck is caused because writing software is an exacting task, somewhat akin to translating "Gone With The Wind" into hieroglyphics. Using special symbolic lann guages (usually COBOL for business applications, and FORTRAN for scientific programs), programmers carefully craft thousands of lines of instructions, which are transformed into magnetic impulses on tape or discs to run the computer.
The basic alphabet of a language is called "systems software," and organizes the computer's electrical impulses into recognizable patterns. "Applications software" then tells the machine what specific tasks it is to undertake.
One misplaced comma or number will hopelessly confuse even the fastest computer, causing it to produce nothing but garbage data.
"Some people view [software design] as engineering," says Paul Gustafson, marketing director for Argonaut Information Systems. "Others think it's an art."
While the technology of integrated circuits has cut the cost of computer hardware a thousandfold since 1948, no similar technological advance has marked the software side of the business. The price of software has doubled since 1953 , boosting the cost of a simple program to $160,000. Programming is fast becoming the major portion of data-processing spending. The US General Accounting Office estimates that by 1985, 90 percent of the federal government's computing expenditures will be for software.
The application programming industry, formed of small companies that peddle stock libraries of software, is growing twice as fast as its hardware counterpart.
Easing the software crunch may depend on good old-fashioned research and development.
"What's going to bring increased productivity in the software into preconcieved patterns, is an attempt at taking the handicraft out of programming and making it an assembly-line procedure.
Intel, a California semi-conductor company, is attempting to put software on its integrated circuits -- in essence installing an automatic translating device on a dime-sized piece of silicon. And the Massachusetts Institute of Technology has developed a new language, named LISP, which needs fewer symbols than COBO or FORTRAN to give computers the same instructions.
All these approaches have their drawbacks. But each is a step toward making programming a simpler procedure.
Eventually industry analysts say, programmers with only a few days of training should be able to instruct computers. highly refined "query languages" may enable users to talk with their computers like they talk to their secretaries.
But these advances are years, if not decades, away.